• No se han encontrado resultados

Presupuesto de implementación de los procesos de comercialización

2 MARCO TEÓRICO

3.8 Comprobación de la hipótesis

5.5.4 Presupuesto de implementación de los procesos de comercialización

Data on the growth of roads, which includes newly built as well as upgraded roads is available from 1988-2006. We intend to use lagged values of road growth in order to eliminate the problem of reverse causality. This constrains us to estimating regression equation (9) only for the years 1993-2006, where the share of gross value added in the year 1993 is regressed on the average of newly built and upgraded roads between 1988-1993, the share of gross value added in 1999 on the average yearly road growth between 1993-1999 and lastly the share of gross value added in the year 2006 on the

growth rate of roads between 1999-2006. The effects of inter-regional infrastructure can be tested for the whole sample, therefore each table includes a specification which leaves out our measure for intra-regional infrastructure in favor of using the full time period from 1988-2006.

Table 4.3: Manufacturing Share and Infrastructure Dependent var.: Regional share of EU15 GVAη= 1.5

(1) (2) (3) (4) (5) exp .541 .613 .823 .770 1.015 (.433) (.419) (.591) (.670) (.479)∗∗ pop. density -.326 -.346 -.332 -.340 -.262 (.158)∗∗ (.156)∗∗ (.142)∗∗ (.136)∗∗ (.116)∗∗ road .014 .012 .015 .014 (.008)∗ (.007)(.007)∗∗ (.007)∗∗ PΛ j6=iφij .738 .742 .719 .469 (.291)∗∗ (.295)∗∗ (.265)∗∗∗ (.233)∗∗ PΛ j6=iφijexpj -.421 -.321 -.386 (.382) (.566) (.434) PΛ

j6=iφijexpj1expj,t>expi,t -.165 -.028

(.367) (.294)

Const. .347 -.432 -.121 -.041 .031

(.173)∗∗ (.285) (.407) (.294) (.242)

Obs. 606 606 606 606 807

R2 .065 .079 .097 .099 .166

Notes: All regressions include time and state fixed effects. The standard errors of each specification are robust against

heteroscedasticity. The sample consists of 195 observations in the first time period and 204 observations over the last three time periods. We miss information on the four French overseas-d´epartements, the three Danish NUTS2 regions and the two autonomous Portuguese regions Madeira and Azores for all time periods. For the nine East-German NUTS2 we miss information only for the first time period.

Table 4.3 presents the results for the manufacturing sector. The first four specifications are constrained to the limited sample from 1993-2006, whereas column five uses the full time span. In each of the first four columns we find a significant positive effect of intra-regional infrastructure. The second specification adds the measure of geographic accessibility which turns out to enter significantly positive too. Regarding the market- access effect and the hub-shadow effect, we do not find significant coefficients in any of the specifications.

Table 4.3 can be summarized as follows: intra-regional infrastructure as well as accessibility affect the regional share of manufacturing industry significantly and positively. One percent growth in roads per square kilometer increases the regional share of manufacturing industry by about 0.015%. The average accessibility of regions manifests in about 0.7% of the regional manufacturing share or put differently, raising the geographic accessibility by 10% above the average results in an increase in the manufacturing share of about 1.1∗0.7 = 0.77%. Yet, the market-access effect and the proximity to larger markets than the own home market seem to play no role for the

manufacturing sector.

Table 4.4: Service Share and Infrastructure Dependent var.: Regional share of EU15 GVAη= 1.5

(1) (2) (3) (4) (5) exp .946 .979 .897 .790 .780 (.127)∗∗∗ (.119)∗∗∗ (.148)∗∗∗ (.157)∗∗∗ (.102)∗∗∗ pop. density .320 .311 .306 .288 .190 (.165)∗ (.164)∗ (.159)∗ (.155)∗ (.122) road .008 .007 .007 .006 (.004)∗∗ (.003)∗∗ (.003)∗∗ (.003)∗∗ PΛ j6=iφij .329 .328 .283 .232 (.119)∗∗∗ (.119)∗∗∗ (.103)∗∗∗ (.108)∗∗ PΛ j6=iφijexpj .164 .367 .311 (.108) (.172)∗∗ (.135)∗∗ PΛ

j6=iφijexpj1expj,t>expi,t -.334 -.339

(.124)∗∗∗ (.105)∗∗∗

Const. -.099 -.446 -.568 -.408 -.251

(.076) (.125)∗∗∗ (.175)∗∗∗ (.139)∗∗∗ (.138)∗

Obs. 606 606 606 606 807

R2 .63 .637 .645 .66 .644

Notes: All regressions include time and state fixed effects. The standard errors of each specification are robust against

heteroscedasticity. The sample consists of 195 observations in the first time period and 204 observations over the last three time periods. We miss information on the four French overseas-d´epartements, the three Danish NUTS2 regions and the two autonomous Portuguese regions Madeira and Azores for all time periods. For the nine East-German NUTS2 we miss information only for the first time period.

The results for the service sector, which are presented in table 4.4, are much more in line with the theory. Again, the first four specifications use the limited sample as they include our measure for intra-regional infrastructure, whereas the last column utilizes the full sample. Just as in the manufacturing sector, the effects of intra-regional infrastructure and geographic accessibility are significantly positive. Yet, both effects are less pronounced. The coefficients for intra-regional infrastructure and geographic accessibility approximately halve in magnitude. In contrast to our results for the manufacturing sector, columns (4) and (5) suggest that the service sector exhibits a significant market-access effect as well as hub-shadow effect. The magnitudes of the coefficients for the market-access effect and the hub-shadow effect are remarkably similar.10 This means that the positive increase in market access is neutralized by the hub-shadow effect if the increase in trade freeness is aimed at a region with a larger home market.

As how to interpret these results, imagine two examples: say a region improves the trade freeness to a region with a smaller home market to ten percent above the average;

10In column (4) the coefficient of the market-access effect is slightly higher than the coefficient of

the hub-shadow effect, while in column (5) it is the inverse. From the 95% confidence interval we cannot reject the equality of the coefficients absolute values in any case.

this results in an increase in the regional service share of about 0.28∗1.1 = 0.308% due to the better geographic accessibility plus about 0.37∗1.1 = 0.407% times the industry share of the other region. However, if the region raises the trade freeness to a region with a larger home market, the total effect will be given by the sum of the two effects already mentioned minus 0.34∗1.1 = 0.374% times the expenditure share of the other region. Hence, perfectly in line with the theory, the hub-shadow effect almost completely cancels out the market-access effect. Still a positive effect from inter-regional infrastructure investment remains through the accessibility channel. Summarizing table 4.4, we can state that the service sector fits all hypotheses from the model. Intra-regional infrastructure turns out to significantly increase the regional share of services, just as the degree of geographic accessibility and the degree of market access do. Moreover, the hub-shadow effect appears to be highly significant and its magnitude relative to the coefficient of the market-access effect is consistent with the theory.